Facebook Graph Search: Why It Will Fail

Jesus Gorriti

Facebook recently released their new search function to great fanfare. They have created a natural language input search box that lets you filter and tap into all the content available to you in your social graph. It looks amazing and seems that it was made to compete directly with Google (or at least drive traffic away from it).

The engineering behind the tool is phenomenal. You can do amazing quick searches that lets you see which of your friends are James Bond fans, find all the photos your family took on that trip to Disneyland, or surface that guy in your friendsphere who works for the cool startup you’re dying to get a job with–all of this using the same language you would use when asking your mother to share her best Italian recipes.


The problem is the cognitive cost of all these searches.

As a user you won’t be tapping into the infinite trove that is the Internet (and where Google searches), but instead into the limited wardrobe you and your friends create. A wardrobe that it is not only limited but that grows slowly as your friends do not add content at the same rate as the rest of the Internet does.

This will limit the variety of results, curbing the enthusiasm for returning to the tool. Imagine that you search for photos of puppies. You are limited to your friends who share photos and who actually have puppies. In Google Image search you might get new results almost every day, but in Facebook, it’s not likely that many of your friends get new puppies that often.


As a user you are forced to create natural language search queries. These queries require a certain amount of cognitive capacity. And for you to know what items are in your inventory. It’s not the same as looking for information in your own bookshelf rather than in a library. In the later, a search function makes sense, in the former, you’ll probably get faster results by browsing. It is true that Facebook has made a huge improvement in natural language recognition, but understanding someone doesn’t guarantee that you are going to get any questions from them. It’s like when you go to a bookstore and you get asked by a helpful assistant what you’re looking for when maybe you’d just like to have a look around.

Most of us would probably have already asked if we knew what we wanted, but when you do not know, when you are just browsing, you do not want to go through the cognitive cost that requires explaining your need to that person. Search is competing with browsing here.


And as a user you don’t want to find out that you have no friends. Since you will be searching in a limited environment, what happens if your search provides zero results? Or results that don’t fit what you were looking for? The actual searches people might invoke do not necessarily match what Facebook intended. Most of our friends actually add little content to Facebook.

There are tons of zombie users who either do not login to Facebook or just don’t care about participating, and prefer to just watch the action. If this is the case in your social graph, then search results are going to be dominated by a very small group of friends (who already dominate you newsfeed). Imagine if you searched Google, but only got results from the same websites.

To sum up why Facebook Graph Search won’t last:

  • Limited inventory of elements to search. This will mean a lot of repeated results, and will cause quick wear on the user’s patience.
  • Complicated queries. The user will need to exert a lot of effort in creating queries, which compete with the relatively low cost of a browsing click.
  • Potential search results might be zero. The worst feeling a user might experience is that of despair; despair produced by the lack of results (or the lack of friends for that matter).

This feature requires too much cognitive cost for the user to really make the most out of it, and competes too closely with many other functionalities that have lower costs. It feels doomed just from the start, and might have been better off as a limited functionality just for photo search, to test and iron out its kinks. It is certainly a bold move.


Of course Facebook was very smart about the examples they demoed the product with, making sure that it looked like an awesome feature. Truth is, most users do not look for those search queries that often, making the demo a concentration of highly infrequent behaviours. And maybe that is where the future of the service lies: identifying those use cases that make most sense to the user, like creating an event or browsing photos, and bringing focus to the search bar only in those scenarios, not trying to just push it to everybody everywhere.

When the search bar reaches a critical mass of use, when Facebook knows there are enough users out there who are going to share their knowledge with other users, then maybe it will be the right moment to start bringing it to the forefront of the user’s experience. Today it still feels like it was a gimmick for investors.

Having said all this, let’s write a caveat: Facebook might just want to mine more data from the users, and this service might improve through time and use. Facebook might not really mind that people don’t use the service that often, happy for it to become another way to keep users engaged just a little bit longer. Time will tell.

Jesus Gorriti

More Stories from Fjord